System general problem solver

Expert system technology was one of the first branches of artificial intelligence [1,2], It is by now well tried and mature. Nevertheless, expert systems are tools to support human thinking, not magic problem-solvers. While the name expert system is based on the notion that the systems behave like experts, it is generally considered that they should be used by experts or at least the well-informed. 

Linear Solvers Although general purpose solvers exist, major improvements in efficiency can be gained by exploiting structural features of the linear equations that are particular to reservoir simulation. The key parts of any linear solver are (i) a method of pre-conditioning, essentially an approximation to the system of equations that can be solved directly, but without storage or speed problems (ii) an iteration scheme. 

The partial differential equations describing the catalyst particle are discretized with central finite difference formulae with respect to the spatial coordinate [50]. Typically, around 10-20 discretization points are enough for the particle. The ordinary differential equations (ODEs) created are solved with respect to time together with the ODEs of the bulk phase. Since the system is stiff, the computer code of Hindmarsh [51] is used as the ODE solver. In general, the simulations progressed without numerical problems. The final values of the rate constants, along with their temperature dependencies, can be obtained with nonlinear regression analysis. The differential equations were solved in situ with the backward... 

The performance of NLP solvers is strongly influenced by the point from which the solution process is started. Points such as the origin (0, 0,...) should be avoided because there may be a number of zero derivatives at that point (as well as problems with infinite values). In general, any point where a substantial number of zero derivatives are possible is undesirable, as is any point where tiny denominator values are possible. Finally, for models of physical processes, the user should avoid starting points that do not represent realistic operating conditions. Such points may cause the solver to move toward points that are stationary points but unacceptable configurations of the physical system. 

Due to these inner iterations via IVP solvers and due to the need to solve an associated nonlinear systems of equations to match the local solutions globally, boundary value problems are generally much harder to solve and take considerably more time than initial value problems. Typically there are between 30 and 120 I VPs to solve numerous times in each successful run of a numerical BVP solver. 

Solution methods for optimization problems that involve only continuous variables can be divided into two broad classes derivative-free methods (e.g., pattern search and stochastic search methods) and derivative-based methods (e.g., barrier function techniques and sequential quadratic programming). Because the optimization problems of concern in RTO are typically of reasonably large scale, must be solved on-line in relatively small amounts of time and derivative-free methods, and generally have much higher computational requirements than derivative-based methods, the solvers contained in most RTO systems use derivative-based techniques. Note that in these solvers the first derivatives are evaluated analytically and the second derivatives are approximated by various updating techniques (e.g., BFGS update). 

In this paper, we shall touch the development of such numerical methods intended for the solution of the coupled evolution problems as e.g. thermoelasticity, which is described in Section 2. Here we also discuss the discretization of the evolution problems. As the computational demands are concentrated mainly in the solution of the arising linear systems, we shall focus on the application of suitable, efficient and parallelizable iterative solvers for these linear systems. Section 3 deals with some general techniques enhancing the efficiency of the iterative solution of discrete evolution problems. Section 4 is devoted to a short discussion of the numerical results. In Section 5, we shall describe solvers, which exploit the domain decomposition and parallel computations. Here we also mention another division techniques as displacement decomposition or composite grid methods. 

---